erl_connect

C LIBRARY

erl_connect

LIBRARY SUMMARY

Communicate with Distributed Erlang

DESCRIPTION

This module provides support for communication between distributed Erlang nodes and C nodes, in a manner that is transparent to Erlang processes.

A C node appears to Erlang as a hidden node. That is, Erlang processes that know the name of the C node are able to communicate with it in a normal manner, but the node name will not appear in the listing provided by the Erlang function nodes/0.

EXPORTS

int erl_connect_init(number, cookie, creation)
int erl_connect_xinit(host, alive, node, addr, cookie, creation)

Types:

int number;

char *cookie;

short creation;

char *host,*alive,*node;

struct in_addr *addr;

These functions initialize the erl_connect module. In particular, they are used to identify the name of the C-node from which they are called. One of these functions must be called before any of the other functions in the erl_connect module are used.

erl_connect_xinit() stores for later use information about the node's host name host, alive name alive, node name node, IP address addr, cookie cookie, and creation number creation. erl_connect_init() provides an alternative interface which does not require as much information from the caller. Instead, erl_connect_init() uses gethostbyname() to obtain default values.

If you use erl_connect_init() your node will have a short name, i.e., it will not be fully qualified. If you need to use fully qualified (a.k.a. long) names, use erl_connect_xinit() instead.

host is the name of the host on which the node is running.

alive is the alivename of the node.

node is the name of the node. The nodename should be of the form alivename@hostname.

addr is the 32-bit IP address of host.

cookie is the authorization string required for access to the remote node. If NULL the user HOME directory is searched for a cookie file .erlang.cookie. The path to the home directory is retrieved from the environment variable HOME on Unix and from the HOMEDRIVE and HOMEPATH variables on Windows. Refer to the auth module for more details.

creation helps identify a particular instance of a C node. In particular, it can help prevent us from receiving messages sent to an earlier process with the same registered name.

A C node acting as a server will be assigned a creation number when it calls erl_publish().

number is used by erl_connect_init() to construct the actual node name. In the second example shown below, "c17@a.DNS.name" will be the resulting node name.

Example 1:

struct in_addr addr;
addr = inet_addr("150.236.14.75");
if (!erl_connect_xinit("chivas",
                       "madonna",
                       "madonna@chivas.du.etx.ericsson.se",
                       &addr;
                       "samplecookiestring..."),
                       0)
  erl_err_quit("<ERROR> when initializing !");

Example 2:

if (!erl_connect_init(17, "samplecookiestring...", 0))
  erl_err_quit("<ERROR> when initializing !");

int erl_connect(node)
int erl_xconnect(addr, alive)

Types:

char *node, *alive;

struct in_addr *addr;

These functions set up a connection to an Erlang node.

erl_xconnect() requires the IP address of the remote host and the alive name of the remote node to be specified. erl_connect() provides an alternative interface, and determines the information from the node name provided.

addr is the 32-bit IP address of the remote host.

alive is the alivename of the remote node.

node is the name of the remote node.

These functions return an open file descriptor on success, or a negative value indicating that an error occurred --- in which case they will set erl_errno to one of:

EHOSTUNREACH: The remote host node is unreachable
ENOMEM: No more memory available.
EIO: I/O error.

Additionally, errno values from socket(2) and connect(2) system calls may be propagated into erl_errno.

#define NODE   "madonna@chivas.du.etx.ericsson.se"
#define ALIVE  "madonna"
#define IP_ADDR "150.236.14.75"

/*** Variant 1 ***/
erl_connect( NODE );

/*** Variant 2 ***/
struct in_addr addr;
addr = inet_addr(IP_ADDR);
erl_xconnect( &addr , ALIVE );

int erl_close_connection(fd)

Types:

int fd;

This function closes an open connection to an Erlang node.

Fd is a file descriptor obtained from erl_connect() or erl_xconnect().

On success, 0 is returned. If the call fails, a non-zero value is returned, and the reason for the error can be obtained with the appropriate platform-dependent call.

int erl_receive(fd, bufp, bufsize)

Types:

int fd;

char *bufp;

int bufsize;

This function receives a message consisting of a sequence of bytes in the Erlang external format.

fd is an open descriptor to an Erlang connection.

bufp is a buffer large enough to hold the expected message.

bufsize indicates the size of bufp.

If a tick occurs, i.e., the Erlang node on the other end of the connection has polled this node to see if it is still alive, the function will return ERL_TICK and no message will be placed in the buffer. Also, erl_errno will be set to EAGAIN.

On success, the message is placed in the specified buffer and the function returns the number of bytes actually read. On failure, the function returns a negative value and will set erl_errno to one of:

EAGAIN: Temporary error: Try again.
EMSGSIZE: Buffer too small.
EIO: I/O error.

int erl_receive_msg(fd, bufp, bufsize, emsg)

Types:

int fd;

unsigned char *bufp;

int bufsize;

ErlMessage *emsg;

This function receives the message into the specified buffer, and decodes into the (ErlMessage *) emsg.

fd is an open descriptor to an Erlang connection.

bufp is a buffer large enough to hold the expected message.

bufsize indicates the size of bufp.

emsg is a pointer to an ErlMessage structure, into which the message will be decoded. ErlMessage is defined as follows:

typedef struct {
  int type;
  ETERM *msg;
  ETERM *to;
  ETERM *from;
  char to_name[MAXREGLEN];
} ErlMessage;

Note

The definition of ErlMessage has changed since earlier versions of Erl_Interface.

type identifies the type of message, one of ERL_SEND, ERL_REG_SEND, ERL_LINK, ERL_UNLINK and ERL_EXIT.

If type contains ERL_SEND this indicates that an ordinary send operation has taken place, and emsg->to contains the Pid of the recipient. If type contains ERL_REG_SEND then a registered send operation took place, and emsg->from contains the Pid of the sender. In both cases, the actual message will be in emsg->msg.

If type contains one of ERL_LINK or ERL_UNLINK, then emsg->to and emsg->from contain the pids of the sender and recipient of the link or unlink. emsg->msg is not used in these cases.

If type contains ERL_EXIT, then this indicates that a link has been broken. In this case, emsg->to and emsg->from contain the pids of the linked processes, and emsg->msg contains the reason for the exit.

Note

It is the caller's responsibility to release the memory pointed to by emsg->msg, emsg->to and emsg->from.

If a tick occurs, i.e., the Erlang node on the other end of the connection has polled this node to see if it is still alive, the function will return ERL_TICK indicating that the tick has been received and responded to, but no message will be placed in the buffer. In this case you should call erl_receive_msg() again.

On success, the function returns ERL_MSG and the Emsg struct will be initialized as described above, or ERL_TICK, in which case no message is returned. On failure, the function returns ERL_ERROR and will set erl_errno to one of:

EMSGSIZE: Buffer too small.
ENOMEM: No more memory available.
EIO: I/O error.

int erl_xreceive_msg(fd, bufpp, bufsizep, emsg)

Types:

int fd;

unsigned char **bufpp;

int *bufsizep;

ErlMessage *emsg;

This function is similar to erl_receive_msg. The difference is that erl_xreceive_msg expects the buffer to have been allocated by malloc, and reallocates it if the received message does not fit into the original buffer. For that reason, both buffer and buffer length are given as pointers - their values may change by the call.

EMSGSIZE: Buffer too small.
ENOMEM: No more memory available.
EIO: I/O error.

int erl_send(fd, to, msg)

Types:

int fd;

ETERM *to, *msg;

This function sends an Erlang term to a process.

fd is an open descriptor to an Erlang connection.

to is an Erlang term containing the Pid of the intended recipient of the message.

msg is the Erlang term to be sent.

The function returns 1 if successful, otherwise 0 --- in which case it will set erl_errno to one of:

EINVAL: Invalid argument: to is not a valid Erlang pid.
ENOMEM: No more memory available.
EIO: I/O error.

int erl_reg_send(fd, to, msg)

Types:

int fd;

char *to;

ETERM *msg;

This function sends an Erlang term to a registered process.

fd is an open descriptor to an Erlang connection.

to is a string containing the registered name of the intended recipient of the message.

msg is the Erlang term to be sent.

The function returns 1 if successful, otherwise 0 --- in which case it will set erl_errno to one of:

ENOMEM: No more memory available.
EIO: I/O error.

ETERM * erl_rpc(fd, mod, fun, args)
int erl_rpc_to(fd, mod, fun, args)
int erl_rpc_from(fd, timeout, emsg)

Types:

int fd, timeout;

char *mod, *fun;

ETERM *args;

ErlMessage *emsg;

These functions support calling Erlang functions on remote nodes. erl_rpc_to() sends an rpc request to a remote node and erl_rpc_from() receives the results of such a call. erl_rpc() combines the functionality of these two functions by sending an rpc request and waiting for the results. See also rpc:call/4.

fd is an open descriptor to an Erlang connection.

timeout is the maximum time (in ms) to wait for results. Specify ERL_NO_TIMEOUT to wait forever. When erl_rpc() calls erl_rpc_from(), the call will never timeout.

mod is the name of the module containing the function to be run on the remote node.

fun is the name of the function to run.

args is an Erlang list, containing the arguments to be passed to the function.

emsg is a message containing the result of the function call.

The actual message returned by the rpc server is a 2-tuple {rex,Reply}. If you are using erl_rpc_from() in your code then this is the message you will need to parse. If you are using erl_rpc() then the tuple itself is parsed for you, and the message returned to your program is the erlang term containing Reply only. Replies to rpc requests are always ERL_SEND messages.

Note

It is the caller's responsibility to free the returned ETERM structure as well as the memory pointed to by emsg->msg and emsg->to.

erl_rpc() returns the remote function's return value (or NULL if it failed). erl_rpc_to() returns 0 on success, and a negative number on failure. erl_rcp_from() returns ERL_MSG when successful (with Emsg now containing the reply tuple), and one of ERL_TICK, ERL_TIMEOUT and ERL_ERROR otherwise. When failing, all three functions set erl_errno to one of:

ENOMEM: No more memory available.
EIO: I/O error.
ETIMEDOUT: Timeout expired.
EAGAIN: Temporary error: Try again.

int erl_publish(port)

Types:

int port;

These functions are used by a server process to register with the local name server epmd, thereby allowing other processes to send messages by using the registered name. Before calling either of these functions, the process should have called bind() and listen() on an open socket.

port is the local name to register, and should be the same as the port number that was previously bound to the socket.

To unregister with epmd, simply close the returned descriptor. See also erl_unpublish().

On success, the functions return a descriptor connecting the calling process to epmd. On failure, they return -1 and set erl_errno to:

EIO: I/O error

Additionally, errno values from socket(2) and connect(2) system calls may be propagated into erl_errno.

int erl_accept(listensock, conp)

Types:

int listensock;

ErlConnect *conp;

This function is used by a server process to accept a connection from a client process.

listensock is an open socket descriptor on which listen() has previously been called.

conp is a pointer to an ErlConnect struct, described as follows:

typedef struct {
  char ipadr[4];             
  char nodename[MAXNODELEN];
} ErlConnect;

On success, conp is filled in with the address and node name of the connecting client and a file descriptor is returned. On failure, ERL_ERROR is returned and erl_errno is set to EIO.

const char * erl_thiscookie()
const char * erl_thisnodename()
const char * erl_thishostname()
const char * erl_thisalivename()
short erl_thiscreation()

These functions can be used to retrieve information about the C Node. These values are initially set with erl_connect_init() or erl_connect_xinit().

int erl_unpublish(alive)

Types:

char *alive;

This function can be called by a process to unregister a specified node name from epmd on the localhost. This may be useful, for example, when epmd has not detected the failure of a node, and will not allow the name to be reused. If you use this function to unregister your own process, be sure to also close the descriptor that was returned by erl_publish().

Note

Careless use of this function may have unpredictable results, if the registered node is in fact still running.

alive is the name of the node to unregister, i.e., the first component of the nodename, without the @hostname.

If the node was successfully unregistered from epmd, the function returns 0. Otherwise, it returns -1 and sets erl_errno is to EIO.

struct hostent *erl_gethostbyname(name)
struct hostent *erl_gethostbyaddr(addr, length, type)
struct hostent *erl_gethostbyname_r(name, hostp, buffer, buflen, h_errnop)
struct hostent *erl_gethostbyaddr_r(addr, length, type, hostp, buffer, buflen, h_errnop)

Types:

const char *name;

const char *addr;

int length;

int type;

struct hostent *hostp;

char *buffer;

int buflen;

int *h_errnop;

These are convenience functions for some common name lookup functions.

Debug Information

If a connection attempt fails, the following can be checked:

erl_errno
that the right cookie was used
that epmd is running
the remote Erlang node on the other side is running the same version of Erlang as the erl_interface library.